Labour Market Policy and Environmental Fiscal Devaluation: A Cure for Spain in the Aftermath of the Great Recession? - Kurt Kratena, Mark Sommer ...

Page created by Hazel Valdez
 
CONTINUE READING
ÖSTERREICHISCHES INSTITUT
FÜR WIRTSCHAFTSFORSCHUNG      WORKING PAPERS

                              Labour Market Policy and
                              Environmental Fiscal Devaluation:
                              A Cure for Spain in the Aftermath
                              of the Great Recession?
                              Kurt Kratena, Mark Sommer

            476/2014
Labour Market Policy and Environmental Fiscal
Devaluation: A Cure for Spain in the Aftermath
of the Great Recession?
Kurt Kratena, Mark Sommer

WIFO Working Papers, No. 476

July 2014

Abstract
This paper evaluates different options of labour market policy and tax reform with payroll tax reductions
for the Spanish economy in the current situation of high unemployment and debt constraints for public
and private households. The Spanish economy in the aftermath of the Great Recession is characterised
by household debt de-leveraging, continuous public spending cuts and stagnation in output and em-
ployment. A disaggregated dynamic New Keynesian (DYNK) model covering 59 industries and five in-
come groups of households is used to evaluate the macroeconomic and labour market impact of the
following policy options: 1. subsidising "green jobs" and reduction of hours worked as an active labour
market policy measure, 2. environmental fiscal devaluation (reductions in social security contributions
balanced by an environmental consumption tax). The results show a significant output and employ-
ment multiplier effect of these policies, given the public budget constraint.

E-mail address: Kurt.Kratena@wifo.ac.at, Mark.Sommer@wifo.ac.at
2014/293/W/0

© 2014 Österreichisches Institut für Wirtschaftsforschung
Medieninhaber (Verleger), Hersteller: Österreichisches Institut für Wirtschaftsforschung • 1030 Wien, Arsenal, Objekt 20 •
Tel. (43 1) 798 26 01-0 • Fax (43 1) 798 93 86 • http://www.wifo.ac.at/ • Verlags- und Herstellungsort: Wien
Die Working Papers geben nicht notwendigerweise die Meinung des WIFO wieder
Kostenloser Download: http://www.wifo.ac.at/wwa/pubid/47399
Labour Market Policy and Environmental Fiscal
    Devaluation: A Cure for Spain in the Aftermath of the Great
                            Recession?

                                     Kurt Kratena, Mark Sommer *

________________________________________________________________________
Abstract:
This paper evaluates different options of labour market policy and tax reform with payroll tax
reductions for the Spanish economy in the current situation of high unemployment and debt
constraints for public and private households. The Spanish economy in the aftermath of the
Great Recesion is characterized by household debt de-leveraging, continuous public spending
cuts and stagnation in output and employment. A disaggregated dynamic New Keynesian
(DYNK) model covering 59 industries and five income groups of households is used to
evaluate the macroeconomic and labour market impact of the following policy options: (i)
subsidizing ‘green jobs’ and reduction of hours worked as an active labour market policy
measure, (ii) environmental fiscal devaluation (reductions in social security contributions
balanced by an environmental consumption tax). The results show a significant output and
employment multiplier effect of these policies, given the public budget constraint.

Key words: New Keynesian model, labour market policy, fiscal devaluation
JEL Code: C54, H23, J68.
________________________________________________________________________
* Austrian Institute of Economic Research (WIFO), Arsenal, Objekt 20, A-1030 Vienna (Austria), corresponding
author: kurt.kratena@wifo.ac.at

Acknowledgments:
This paper is dedicated to Cristina Guerrero, who unrestingly attempts to make economists aware of the current
social situation in Spain. Invaluable research assistance has been provided by Katharina Köberl.
2

Introduction

The economic situation of the European periphery countries (Greece, Ireland, Portugal and
Spain) in the aftermath of the Great Recession 2009 and the turmoil on European financial
markets for public debt is characterized by high public and private (household) debt and high
unemployment rates. Austerity policies have been introduced in these countries in order to
reduce public deficits and the public debt burden. The experience of several years of missing
these targets for public debt reduction has in turn opened a discussion on the correct
estimation of fiscal multipliers. The body of literature on this issue has been growing very fast
and has dealt with several macroeconomic aspects that are relevant for the magnitude of fiscal
multipliers, see Illing and Watzka (2014) for a recent example. In model simulations In't
Veld, J., (2013) has shown that the inter-regional spillover effects of simultaneous austerity
policies among all European member states might be significant. The debate is not closed, but
there is also some evidence that though fiscal multipliers are not constant and might be
condiderably larger in times of financial crisis (Müller, 2014), austerity policy might not
necessarily be self defeating, i.e. missing its own targets due to macroeconomic feedback
effects.
On the other hand, as Blanchard, et al. (2013) have shown, in the aftermath of the Great
Recession job creation might stay relatively low for some period, leading to sustained high
unemployment rates accompanied by social problems. Therefore the question about policy
options for lowering unemployment and fostering GDP growth, given the public debt and
deficit constraints, arises. This is also the main issue of this paper, exemplified for the case of
the Spanish economy. These policy options can be discussed from the perspective of their
short-term contribution to alleviating the post-crisis economic and social problems, but at the
same time also from their compatibility with mid-term strategic targets of European social
and environmental policy. Ideally, improving the labour market and economic situation in the
short-run, would not be counterproductive for achieving the mid-term targets.
This paper deals with two main policy options that can be implemented without violating the
public debt-targets: (i) active labour market policies, and (ii) environmental fiscal
devaluation. The latter has been intensively discussed for countries that have serious public
budget constraints, see Farhi, et al. (2014) and De Mooji and Keen (2012), among others.
3

Fiscal devaluation can be seen as a special version of internal devaluation, where the decrease
in wage costs and the increase in competitiveness are achieved by a shift in taxation from
labour to consumption. The impact on domestic prices in foreign currency of this taxation
shift is similar to the impact of currency devaluation. Farhi, et al. (2014) have shown how this
positive impact on the domestic economy depends on feedback effects from the labour
market, especially wage setting. For Spain, Alvarez-Martinez and Polo (2014) have also
analysed other potential tax increases to balance the decrease in labour taxation, like income
taxation and a decrease in the unemployment benefit rate. The latter is thought to counteract
labour fraud and the shadow economy. Simulations of green taxation for Spain (Markandya,
et al., 2013) have shown that assumptions about the magnitude and the reaction of the shadow
economy play an important role in quantifying the welfare effect of tax reforms. The study by
CPB (2013) shows for several European countries the results of model simulation based on
macroeconomic models for normalized fiscal devaluations. In general, fiscal devaluation is
well researched in a theoretical context or in macroeconomic model simulations. The model
simulations presented here build on that work and add important industry details and combine
environmental and competitiveness considerations.
The paper is organized as follows. The first section describes the main blocks of the model:
household behavior and private consumption, firm behavior and production structure, the
labour market, and the government sector including model closure. In the second section the
data bases used, the main econometric estimation results for a panel of EU countries and the
calibration methodology of the model for the Spanish economy are described. Finally, the
third section contains the design and the results of the policy simulations with the model. The
first simulation describes the impact of active labour market policies for Spain, where hiring
subsidies for unemployed workers are used to match these unemployed persons with green
jobs in recycling activities and refurbishing of buildings. This measure is accompanied by a
shortening of the working time, where part of the income loss is compensated by public
transfers, similar to the German "Kurzarbeit" sheme. Besides the labour market impact, these
policies also have an impact on direct and indirect energy use of households and therefore on
GHG emissions. The other simulation describes the results of 'environmental fiscal
devaluation'. This fiscal devaluation is designed as a consumption tax on embodied GHG
emissions in each commodity (according to the the 59 industry classification) with tax rates
4

according to the EU Low Carbon Roadmap 2050 (European Commission, 2011) which is
compensated by lower social security contributions by both employers and employees. The
main idea behind this tax reform is combining potential short-run positive impacts on
employment with long-run environmental targets. A crucial issue for this type of
environmental fiscal devaluation is competitiveness: the consumption tax reduces internal
demand and does not harm (price) competitiveness of the domestic industry like a general tax
on GHG emissions (including production). Additionally, this consumption GHG emission tax
does not lead to carbon leakage, as imported embodied emissions are levied in equal
magnitude as domestic embodied emissions.

1. The model
The model approach applied can be characterized as a DYNK (DYnamic New Keynesian)
model with rigidities and institutional frictions. In that aspect, the DYNK model bears some
similarities with the DSGE (Dynamic Stochastic General Equilibrium) approach which has
been intensively used recently in the analysis of labour market policy (Busl and Seymen,
2013 and Faia, et al., 2012). The model explicitly describes an adjustment path towards a
long-run equilibrium. This feature of dynamic adjustment towards equilibrium is most
developed in the consumption block and in the macroeconomic closure via a fixed short and
long-term path for the public deficit. The term ‘New Keynesian’ refers to the existence of a
log-run full employment equilibrium, which will not be reached in the short run, due to
institutional rigidities. These rigidities include liquidity constraints for consumers (deviation
from the permanent income hypothesis), wage bargaining (deviation from the competitive
labour market) and an imperfect capital market. Depending on the magnitude of the distance
to the long-run equilibrium, the reaction of macroeconomic aggregates to policy shocks can
differ substantially.
The model describes the inter-linkages between 59 industries as well as the consumption of
five household income groups by 47 consumption categories. The model is closed by
endogenizing parts of public expenditure in order to meet the mid-term stability program for
public finances in Spain.

1.1 Household behaviour and private consumption
5

The consumption decision of households in the DYNK model is modeled along the lines of
the ‘buffer stock model’ of consumption (Carroll, 1997), including consumption of durables
and nondurables (Luengo-Prado, 2006).
Durable demand and total nondurables
Consumers maximize the present discounted value of expected utility from consumption of
nondurable commodity and from the service provided by the stocks of durable commodity:

                     ∞                 
        max V = E 0 ∑ β tU (C t , K t )                                                (1)
        ( Ct , K t )
                      t =0             
Specifying a CRRA utility function yields:

                         C t1− ρ    K t1− ρ
        U (C t , K t ) =         +ϕ                                                      (2)
                         1− ρ       1− ρ

where ϕ is a preference parameter and ρ > 0 implies risk aversion of consumers.

The budget constraint in this model without adjustment costs for the durables stock is given
by the definition of assets, At:

        At = (1 + r )(1 − tr ) At −1 + YDt − Ct − (K t − (1 − δ )K t −1 )                (3)

In (3) the sum of Ct and (K t − (1 − δ )K t −1 ) represents total consumption, i.e. the sum of

nondurable and durable expenditure (with depreciation rate of the durable stock, δ). The gross
profit income rAt-1 is taxed with tax rate tr. These taxes therefore reduce the flow of net
lending of households that accumulates to future assets. Disposable household income
excluding profit income, YDt, is given as the balance of net wages (1 − t S − tY )wt H t and net

operating surplus accruing to households (1 − tY )Π h ,t , plus unemployment benefits transfers

with UNt as unemployed persons and br as the benefit replacement rate, measured in terms of
the after tax wage rate, plus other transfers Trt:

        YDt = (1 − t S − tY )wt H t + (1 − tY )Π h ,t + brwt (1 − t S − tY )UN t + Trt   (4)
The following taxes are charged on household income: social security contributions with tax
rate tS, which can be further decomposed into an employee and an employer’s tax rate (twL and
tL) and income taxes with tax rate tY. The wage rate wt is the wage per hour and Ht are total
6

hours demanded by firms. Wage bargaining between firms and unions takes place over the
employee’s gross wage, i.e. wt (1 - tL).
Financial assets of households are built up by saving after durable purchasing has been
financed, and the constraint for lending is:

          At + (1 − θ )K t ≥ 0                                                              (5)

This term represents voluntary equity holding, Qt+1 = At + (1 - θ)Kt, as the equivalent of the
other part of the durable stock (θKt) needs to be held as equity. The consideration of the
collateralized constraint is operationalized in a down payment requirement parameter θ,
which represents the fraction of durables purchases that a household is not allowed to finance.
One main variable in the buffer stock-model of consumption is ‘cash on hand’, Xt, measuring
the household’s total resources:

       Xt = (1 + rt)(1 – tr)At-1 + (1 - δ)Kt-1 + YDt                                        (6)

Total consumption is then defined as:

       CPt = Ct + Kt - (1 - δ)Kt-1 = rt(1 – tr)At-1 + YDt – (At-1 - At)                     (7)

In (7) the last term represents net lending, so total consumption is the sum of durable and
nondurable consumption or the difference between disposable income and net lending.

                                                                          ∂U t     ∂U t
The model solution works via deriving the first order conditions for           and      taking into
                                                                          ∂C t     ∂K t

           ∂C t
account         . Luengo-Prado (2006) arrives at an intra-temporal equilibrium relationship
           ∂K t
between Ct and Kt (mostly following Chah, et al., 1995) as one solution of the model, where
the constraint is not binding, or (which is equivalent) the down payment share θ equals the
             r +δ
user costs        . For all other cases, where the collateral constraint is binding, Luengo-Prado
             1+ r
(2006) has shown that this relationship can be used to derive policy functions for Ct and Kt
and formulate both as functions of the difference between cash on hand and the equity that the
consumer wants to hold in the next period. A non-linear consumption function for durables,
similar to the function described in Luengo-Prado and Sørensen (2004) for nondurables, is
asuumed, stating that consumers seek for an equilibrium relationship of durables per
7

household, h.          Therefore, with higher levels of durables per households, the marginal
propensity of investment in durables, CKt with respect to Xt decreases (βK,4 < 0) according to:

          log C dur ,t = β K + β K ,1 log X t + β K , 2θ Ct + β K ,3 log( pdur ,t (rt + δ ) )
                                                                                                                             (8)
          + β K , 4 log X t log(K t −1 / ht −1 )

Note that Cdur,t is equal to Kt - (1 - δ)Kt-1 in equation (7). The down payemt parameter θ in
Luengo-Prado (2006) represents a long-term constraint between the liabilities stock and the
durable stock of households that is imposed on financial markets and might change over time.
Changes in this constraint on stocks can only be achieved in the long-term by imposing limits
to the down payment for durable purchases, θCt.

Equation (8) can be seen as the long-run relationship between Cdur,t and Xt. The long-run
marginal propensity of durable demand to cash on hand depends on the accumulated stock
Kt/ht and is defined by: β K ,1 + β K , 4 log(K t −1 / ht −1 ) In the long-run, with rising income,
                                                              .
households do not keep the relationship between durables and income constant, but the
relationship between voluntary equity holding and income. That corresponds to the long-run
solution of the buffer stock model without durables, where all equity accumulation is
voluntary, because no collateral constraint is active. Usually, in the buffer stock model, non-
stationarity of consumption, income and wealth is dealt with by normalizing the variables by
dividing through permanent income. In this paper, instead, the non-stationarity is taken into
account by formulating adjustment processes of short-term behavior towards long-run optimal
relationships. Therrefore, demand for durables is formulated as an error correction mechanism
(ECM), like in Caballero, 1993 and Eberly, 1994:

                                                        log C dur ,t −1 − β K + β K ,1 log X t −1 + β K , 2θ C ,t −1 +                  
d log C dur ,t = γ K + γ K ,1 log d log X t + γ K , ECM                                                                                 
                                                         β K ,3 log( pdur ,t −1 (rt −1 + δ ) ) + β K , 4 log X t −1 log(K t −2 / ht −2 )
                                                                                                                            (9)

In (9) βK and γK are constants (in the panel data regression cross section fixed effects), and
γK,ECM represents the ECM parameter with γK,ECM < 0. Equation (9) is specified for own
houses (dwelling investment) and for vehicles (Chous,t and Cveh,t). The capital stock for both
durables categories (Khous,t and Kveh,t) accumulates according to the following equation:
K t = K t −1 (1 − δ ) + C dur ,t starting from an estimated initial durable stock in t = 0. The

depreciation rates (δ) are specific for both durable categories. Durable consumption is in
8

equation (7) described as an investment (Kt - (1 - δ)Kt-1), which is the case for one of the two
durable categories, namely expenditure for vehicle purchases. For own houses the
consumption data do not contain dwelling investment for own houses, but imputed rents. This
is due to the concepts in national accounting, which treat housing different from other
durables. The imputed rents are calculated as a simple static user cost: C rent ,t = pdur ,t (rt + δ )K t .

The demand function for total nondurable consumption is modeled with a positive marginal
propensity of nondurable consumption to ‘cash on hand’ and a negative marginal propensity
of total nondurable consumption to the product of the down payment (in percentage of
durables) and durable demand:

          log Ct = β C + β C ,1 log X t + β C , 2θ Ct log C dur ,t                                             (10)

This function takes into account that households need to finance the sum of Ct + θ Ct C dur ,t , but

down payments will not be fully financed by savings in the same period and consumers
smooth nondurable consumption accordingly. This smoothing is measured in (10) by the
parameter βC,2.

The long-run marginal propensity of nondurable demand to cash on hand is given by the
direct impact (βC,1) plus the indirect impact via θCtlogCdur,t . The latter again depends on
β K ,1 + β K , 4 log(K t −1 / ht −1 ) , so that the total marginal propensity of nondurable demand to cash

on hand is defined by: β C ,1 + θ Ct β C , 2 β K ,1 + θ Ct β C , 2 β K , 4 log(K t −1 / ht −1 )

The second term in this relationship measures the necessary increase in savings for down
payments due to an increase in durable demand, induced by a marginal increase in cash on
hand. The last term measures the impact of the non-linearity in the reaction of durables
demand to cash on hand on savings (and on nondurable demand). Note that as durable
demand reacts to the price of durables and nondurable demand is linked to durable demand in
(10), there is also an implicit price elasticity for nondurables at work. Like in the case of
durable demand, the error correction mechanism (ECM) representation of (10) is:

d log Ct = γ C + γ C ,1 log d log X t + γ C ,ECM [log Ct −1 − β C + β C ,1 log X t −1 + β C , 2θ C ,t −1 log Cdur ,t −1 ]
                                                                                                                            (11)
Energy demand
9

The energy demand of households comprises fuel for transport, electricity and heating. These
demands are part of total nondurable consumption and are modeled in single equations,
therefore assuming separability from non-energy nondurable consumption. According to the
literature on the rebound effect (e.g.: Khazzoom, 1989), the energy demand is modeled as
(nominal) service demand and the service aspect is taken into account by dealing with service
prices. The durable stock of households (vehicles, houses, appliances) embodies the
efficiency of converting an energy flow into a service level S = ηES E, where E is the energy
demand for a certain fuel and S is the demand for a service inversely linked by the efficiency
parameter (ηES) of converting the corresponding fuel into a certain service. For a given
conversion efficiency, a service price, pS, (marginal cost of service) can be derived, which is a
function of the energy price and the efficiency parameter. Any increase in efficiency leads to
a decrease in the service price and thereby to an increase in service demand ('rebound effect').

                pE
        pS =
               η ES                                                                       (12)
For transport demand of households we take substitution between public (CPpub) and private
transport (CPfuel) into account. For this purpose, a price (pctr) of the aggregate transport
demand, CPtr, is constructed:

                    CPfuel                  CPpub            
        pctr = exp         log pcS , fuel +       log pc pub                            (13)
                    CPtr                    CPtr             

The price for fuels, pcS,fuel , is defined as a service price. Total transport demand of
households depends on this aggregate price as well as on total nondurable expenditure in a
log-linear specification, so that the price and expenditure elasticity can be derived directly
from the parameters (βtr,1 and βtr,2) :

        log CPtr = µ tr + β tr ,1 log pctr + β tr , 2 log Ct                              (14)

In (14) µtr is a constant or a cross section fixed effect in the panel data model.

The demand for transport fuels is linked to the vehicle stock and depends on the service price
of fuels as well as on the endowment of vehicles of the population. The latter term is
important because the second car of the household usually is used less in terms of miles
driven than the first.
10

               CPfuel ,t                         p                       K        
          log              = µ fuel + γ fuel log fuel ,t    + ξ fuel log veh,t      (15)
                                                  η                        h
               K veh,t                            fuel ,t                 t        

In (15) µfuel again is a constant or a cross section fixed effect and γfuel is the price elasticity
under the condition that there is a unitary elasticity of fuel demand to the vehicle stock. Once
total transport demand of households and demand for fuels for transport are determined,
public transport demand can be derived as a residual.
The equations for heating and electricity demand are analogous to equation (15) and have the
following form:

               CP                                p          
          log heat ,t      = µ heat + γ heat log heat ,t    + ξ heat log(dd heat )     (16)
                                                  η          
               K hous ,t                          heat ,t   

              CP                         p        
          log el ,t     = µ el + γ el log el ,t    + ξ el log(dd heat )                 (17)
             K                           η        
              app ,t                      el ,t   
In both equations the variable heating degree days ddheat is added. All equations also contain
autoregressive terms that have been omitted in this presentation. The durable stocks used are
the total housing stock (Khous,t) and the appliance stock (Kapp,t). The latter is accumulated from
consumption of appliances, CPapp, which in turn is explained in a log linear specification like
total transport demand:

          log CPapp = µ app + β app ,1 log pcapp + β app , 2 log Ct                         (18)

Again, µapp is a constant or a cross section fixed effect. The total housing stock (Khous,t)
contains the stock of own houses, which is explained above and the stock of houses that are
rented by households. The latter is driven by population dynamics.
Nondurable (non-energy) demand
The non-energy demand of nondurables is treated in a demand system. The one applied in this
DYNK model is the Almost Ideal Demand System (AIDS), starting from the cost function for
C(u, pi), describing the expenditure function (for C) as a function of a given level of utility u
and prices of consumer goods, pi (see: Deaton and Muellbauer, 1980) . The AIDS model is
represented by the well known budget share equations for the i nondurable goods in each
period:
11

                                             C 
          wi = α i + ∑ γ ij log p j + β i log  ; i = 1...n, 1...k                                                  (18)
                     j                       P

with price index, Pt, defined by log Pt = α 0 + ∑ α i log pit + 0.5∑∑ γ ij log pit log p jt , often
                                                                   i                    i    j

approached by the Stone price index: log Pt = ∑ wit log pit .
                                                          *

                                                                   k

The expressions for expenditure (ηi) and compensated price elasticities ( ε ijC ) within the AIDS

model for the quantity of each consumption category Ci can be written as (the details of the
derivation can be found in Green, and Alston, 1990) 1:

                    ∂ log C i β i
          ηi =               =    +1
                    ∂ log C wi                                                                                        (19)

                    ∂ log C i γ ij − β i w j
          ε ijC =             =              − δ ij + ε i w j
                    ∂ log p j       wi
                                                                                                                      (20)

In (20) δij is the Kronecker delta with δij = 0 for i ≠ j and δij = 1 for i = j.

The commodity classification i = 1...n in this model comprises the n non-energy nondurables:
(i) food, and beverages, tobacco, (ii) clothing, and footwear, (iii) furniture and household
equipment, (iv) health, (v) communication, (vi) recreation and accomodation, (vii) financial
services, and (viii) other commodities and services.
Total household demand
The household model described determines in three stages the demand for different categories
of durables, energy demand and different categories of nondurables. The vector of non-energy
nondurable consumption, (c NE ) as described above, is given by multiplying total non-energy
nondurable expenditure C with the column vector of budget shares, w (all bold characters are
vectors or matrices), as determined in (18):

          (c NE ) = Cw                                                                                                (21)

    The total consumption vector of categories of consumption in National Accounts (according
to the COICOP classification), cC, is transformed into a consumption vector by commodities

1The derivation oft the budget share wi with respect to log (C) and log (pj) is given by βi and γij – βi (log(P)) respectively.
Applying Shephard’s Lemma and using the Stone price approximation, the elasticity formulae can then be derived.
12

of the input-output core in the DYNK model in purchaser prices, cpp, by applying the bridge
matrix, BC:

                                     ..               ..           .. 
                                                                   
cpp = BC cC ; cC =            cNE =  c i  ; cE =      cf  ;   k = c j                     (22)
                                     ..               ..           .. 
                                                                   

where i = 1...n, f = 1...k, and j = 1...m.
The bridge matrix links the vectors and has the dimension industries in NACE classification *
consumption categories in COICOP classification. Multiplying the vector cC in equation (22)
                                                                                     [
by the bridge matrix BC and by a diagonal matrix of import shares C m or by I − C m with I  ]
as the identity matrix, yields the vector of imported consumption goods ( c mpp ) and of

consumption goods from domestic production ( c dpp ) respectively, both in purchaser prices:

                         c NE                            c NE 
        c   m
            pp
                   ˆ m
                        
                 = C BC  cE 
                               
                                      c   d
                                          pp    [    ˆ m
                                                        ] 
                                               = I − C BC  cE 
                                                                 
                         k                               k 
                                                                                            (23)
After this multiplication, in a first step, taxes less subsidies are subtracted in order to arrive at
consumption vectors net of taxes, c mN and c dN :

                  [ˆ cm
        c mN = I − TN pp  ]                     [ˆ cd
                                      c dN = I − TN pp ]                                        (24)

In (24), T̂N is a diagonal matrix of net tax rates (with identical tax rates on domestic and
imported commodities), and total net taxes (taxes less subsidies) from consumption are
therefore given by:
             ˆ cm + cd
        TN = TN pp    [
                     pp          ]                                                              (25)

By subtracting trade and transport margins as well, we arrive at consumption vectors in basic
prices that determine consumption demand by detailed commodity.
13

1.2 Firm behaviour and production structure
The production side in the DYNK model is analysed within the cost and factor demand
function framework, i.e. the dual model, in a Translog specification. The representative
producers in each industry all face a unit cost function with constant returns to scale that
determines the output price (unit cost), for given input prices. The input quantities follow
from the factor demand functions, once all prices are determined. The Translog specification
chosen in the DYNK model comprises different components of technological change.
Autonomous technical change can be found for all input factors (i.e. the factor biases) and
also as the driver of TFP (total factor productivity), measured by a linear and a quadratic
component.
Substitution in a K,L,E,Mm,Md model
The Translog model is set up with inputs of capital (K), labor (L), energy (E), imported (Mm)
and domestic non-energy materials (Md), and their corresponding input prices p K , p L , p E ,

p Mm and p Md . Applying Shepard’s Lemma yields the cost share equations in the Translog
case, which in turn are used to derive the quantities of factor demand for (K), (L), (E), (Mm)
and (Md). For this production system the input prices can be viewed as exogenous. One part of
the input prices is determined at national or global factor markets, which applies to the prices
of (K), (L), and (E). The price of labour is determined at the labour market via wage functions
by industry (see below). The price of capital is formulated as a simple static user cost price
index with the following components: (i) the price of investment by industry, (ii) the
smoothed interest rate, and (iii) the fixed depreciation rate. The financial market and monetary
policy are not described in detail in the DYNK model, therefore the interest rate is assumed as
exogenous and is approximated by the smoothed benchmark interest rate. The depreciation
rate by industry is fixed (see below for data sources) and the price of investment by industry
is endogenously derived from the price system in the DYNK model. The price of energy
carriers is assumed to be determined at world markets for energy and is therefore treated as
exogenous. Each industry faces a unit cost function for the price (pQ) of output Q, with
constant returns to scale

                                   1                                                             1
log pQ = α 0 + ∑ α i log( pi ) +     ∑ γ ii (log( pi ) )2 + ∑ γ ij log( pi ) log( p j ) + α t t + α tt t 2 + ∑ ρ ti t log( pi )
                i                  2 i                      i, j                                 2            i

                                                                                                             (26)
14

, where pQ is the output price (unit cost), pi, pj are the input prices for input quantities xi, xj,
and t is the deterministic time trend. Note that equation (24) comprises different components
of technological change. Autonomous technical change can be found for all input factors (i.e.
the factor biases, ρ ti ). Another source of autonomous technical change that only influences

unit costs is TFP, measured by α t , and α tt .

The Translog model is set up with inputs of capital (K), labor (L), energy (E), imported (Mm)
and domestic non-energy materials (Md), and their corresponding input prices p K , p L , p E ,

p Mm and p Md . As is well known, Shepard’s Lemma yields the cost share equations in the
Translog case, which in this case of five inputs can be written as:

v K = [α K + γ KK log( p K / p Md ) + γ KL log( p L / p Md ) + γ KE log( p E / p Md ) + γ KM log( p Mm / p Md ) + ρ tK t ]
v L = [α L + γ LL log( p L / p Md ) + γ KL log( p K / p Md ) + γ LE log( p E / p Md ) + γ LM log( p Mm / p Md ) + ρ tL t ]
v E = [α E + γ EE log( p E / p Md ) + γ KE log( p K / p Md ) + γ LE log( p L / p Md ) + γ EM log( p Mm / p Md ) + ρ tE t ]
v M = [α M + γ MM log( p Mm / p Md ) + γ KM log( p K / p Md ) + γ LM log( p L / p Md ) + γ EM log( p E / p Md ) + ρ tM t ]
                                                                                                   (27)

The homogeneity restriction for the price parameters                 ∑γ
                                                                      i
                                                                          ij   = 0,   ∑γ
                                                                                      j
                                                                                           ij   = 0 has already been

imposed in (27), so that the terms for the price of domestic intermediates p Md have been
omitted. In this model, labour and import demand react to changes in the prices of all inputs
and changes in time due to the factor biases that can be labour saving or labour using, as well
as import saving or import using. The immediate reaction to price changes is given by the
own and cross price elasticities. These own- and cross- price elasticities for changes in input
quantity xi can be derived directly, or via the Allen elasticities of substitution (AES), and are
given as:

                ∂ log xi vi2 − vi + γ ii
         ε ii =          =                                                                                     (28)
                ∂ log pi       vi

                  ∂ log xi    vi v j + γ ij
         ε ij =             =                                                                                  (29)
                  ∂ log p j         vi

Here, the vi represent the factor shares in equation (27), and the γij the cross-price parameters.
15

The deterministic trend t captures the two different sources of autonomous technological
change that together influence factor demand, i.e. TFP and the factor bias. The total impact of
t on factor xi is given by:

          d log xi ρ ti
                  =     + α t + α tt t                                                                                (30)
             dt     vi

This impact therefore depends negatively on the share and on the level of technology, due to
the term αttt. The factor shares vi in (27) can be directly used to derive factor demand (in
nominal terms), once the output at current prices pQQ is given. For given input prices p L , p E ,

p Mm and p Md this can be transformed into factor demand in real terms (hours worked or
employees for L and physical energy units for E). A special treatment is applied to the capital
input. This is due to the inherent difference between the ex post rate of return to K that is
implicit in treating operating surplus as the residual in total output and the ex ante rate of
return to K used for the specification of the price of K (user cost). In economic terms, that
represents an imperfect capital market, which can be in disequilibrium (see: Jorgenson, et al.,
2013) so that the adjustment of the ex post rate of return to K towards the ex ante rate of
return to K takes time. This adjustment is specified as a simple ECM process in each industry:

d log( p K ,t ) = α K + β K d log[ pCF ,t (rt + δ )] + τ K [log( p K ,t −1 ) − µ K − γ K [ pCF ,t −1 (rt −1 + δ )]]   (31)

The adjustment towards equilibrium is guaranteed by τK < 0, r is the smoothed (MA process)
benchmark interest rate and δ is the industry specific depreciation rate. The price pCF is the
price of investment (capital formation) by industry and both αK and µK are cross section fixed
effects across countries. Once pK is determined, the factor share for K in (27) can be used to
determine Kjt by industry (j), which in turn determines investment by industry CFj by
inverting the capital accumulation equation:

          CF j ,t = K j ,t − (1 − δ )K j ,t −1                                                                        (32)

Intermediate input demand and factor prices
The factors E, Mm, and Md are aggregates of the use matrix from the supply and use table
system, which is the framework of this DYNK model. The aggregate E comprises four energy
industries/commodities, and Mm, Md the other 55 non-energy industries/commodities (see the
Appendix for the full 59 industry/commodity classification of the DYNK model).
16

In a second nest, the factor E is split up into aggregate categories of energy (coal, oil, gas,
renewable, electricity/heat) in a Translog model. The unit cost function of this model
determines the bundle price of energy, p E , and the cost shares of the five aggregate energy
types:

                                               1
log pE = α 0 + ∑ α E ,i log( pE ,i ) +           ∑ γ E ,ii (log( pE ,i ) )2 + ∑ γ E ,ij log( pi ) log( p j ) + ∑ ρtE ,i t log( pE ,i )
                      i                        2 i                            i, j                             i

                                                                                                                            (33)

                                                   
vE ,i = α E ,i + ∑ γ E ,ij log( pE ,i ) + ρtE ,i t                                                                         (34)
                 i, j                              

This set of energy categories is directly linked to the energy commodities/industries of the use
table.

The domestic as well as the import matrix are converted into ‘use structure matrices’ S mNE and

S dNE by dividing by the column sum of total domestic and imported non-energy intermediates,
respectively. Intermediate inputs by commodity are determined by multiplying diagonal
matrices of the factor shares in (27), V̂D and V̂M with the ‘use structure matrices’ and with
the column vector of output in current prices. The full commodity balance is given by adding
the column vector of domestic consumption (equation (24)), capital formation by domestic
goods, and other domestic final demand (exports exd, changes in stocks std and public
consumption cgd). The capital formation vector by domestic goods is derived by multiplying
the vector of investment by industry cfj (equation (32)) with the capital structure matrix for
investment, derived from the capital formation matrix (investment by industry * investment
by commodity) for domestic investment demand: cf d = B dK cf j . The total investment structure

matrix is made up of domestic and imported investment structures ( B dK and B mK ) and has
column sum of one like the private consumption bridge matrix.

The (column vector) of the domestic output of commodities in current prices, p DQ D , is
transformed into the (column vector) of output in current prices, p QQ , by applying the

market shares matrix, C (industries * commodities) with column sum equal to one:

           p DQ D = VD NE [Q      ]
                    ˆ S d p Q + c d + cf d + ex d + st d + cg d                                                              (35)
17

        p QQ = Cp DQ D                                                                           (36)

The final demand categories in (35), i.e. cd, cfd, exd, std and cgd are all in current prices.
Factor prices are exogenous for the derivation of factor demand, but are endogenous in the
system of supply and demand. Some factor prices are directly linked to the output prices pQ
which are determined in the same system. All user prices are the weighted sum of the
domestic price pd and the import price, pm. The import price of commodity i in country s is
given as the weighted sum of the commodity prices of the k sending countries ( p d ,k )
                s −1
        pim, s = ∑ wmk , s p d ,k                                                                (37)
                k =1

This is derived from an inter-regional input-output system from the WIOD database (see next
section). This gives one domestic price per user for each commodity (i.e. no price
differentiation for domestic goods) and different import prices per user for each commodity,
given by the different country source structure of imports of the same commodity by user.
Once this user specific prices for intermediate goods are given, the ‘use structure matrices’ (
S mNE and S dNE ) can be applied in order to derive the price vectos pMm and pMd:

        p Mm = p mS mNE              p Md = p d S dNE                                            (38)

The price of capital is based on the user cost of capital: u K = pCF (r + δ ) with pCF as the price

of investment goods an industry is buying, r as the deflated benchmark interest rate and δ as
the aggregate depreciation rate of the capital stock K. The investment goods price pCF can be
defined as a function of the domestic commodity prices and import prices, given the input
structures for investment, derived from the capital formation matrix described above for
domestic and imported investment demand:

        p CF = p m B mK + p d B dK                                                               (39)

It is important to note that by these input-output loops in the model, indirect effects or
feedback effects of prices occur and factor demand reactions therefore differ from what the
ceteris paribus price and substitution elasticities indicate. All user prices (for example the
price of private consumption) can further be aggregated in order to derive the aggregate price
index of the corresponding demand aggregate.
18

1.3 Labour market
The main factor market that has important repercussions in the case of policy simulations in
the DYNK model is the labour market. In CGE modeling, different labour market approaches
can be integrated (Boeters and Savard, 2013) and calibrated. In this exercise, the theoretical
approaches need to be confronted with the results from empirical wage curve estimation,
which can be seen as a robust empirical relationship (Card, 1995 and Blanchower and
Oswald, 1994). The wage curves in the DYNK model are specified as the employee’s gross
wage rate per hour by industry, i.e. wt (1 - tL). The labour price (index) of the Translog model
is then defined by adding the employers' social security contribution to that. Combining the
meta-analysis of Folmer (2009) on the empirical wage curve literature with a basic wage
bargaining model from Boeters and Savard (2013) gives a specification for the sectoral hourly
wages. These functions describe the responsiveness of hourly wages to labour productivity
(industry, aggregate), consumer prices, hours worked per employee, and the rate of
unemployment. The inclusion of the variable ' hours worked per employee' corresponds to a
bargaining model, where firms and workers (or unions) bargain over wages and hours worked
simultaneously (Busl and Seymen, 2013). The basic idea is that the gains in labour
productivity can be used for cutting hours worked and wage increases simultaneously. While
unions formally bargain over an hourly wage rate, they also take annual (or monthly) wage
income per head into account (for example for minimum wage considerations). We specify
the wage function in a way that the hours can be determined in a first step and then the hourly
wage rate is determined. A bargaining over hours that leads to less hours worked, would
ceteris paribus lower annual wage income per head. Therefore unions, in consequence,
bargain an increase in the hourly wage rate, so income per year does not fall in the
proportional amount of working time reduction. This specification follows the assumption
that the productivity increase is never fully compensated only by a reduction of working time,
but split up into working time reduction and wage increase. The parameter estimated for
labour productivity in the wage curve therefore is conditional on this impact of working time
on hourly wages.
In the search model firms and workers bargain over the distribution of the value of a
successful match and the wage rate can be derived from the optimality conditions of the
problem (see: Boeters and Savard, 2013):
19

           (1 − t )λ  ρ + urs 
                         m

        w=                      γ                                                                                                 (40)
            π (1 − br )(1 − t )
                     V
                                       a

In (40) tm and ta represent marginal vs. average income tax rates, therefore, if we assume a
proportional tax system for simplicity (and for the sake of data availability) this wage
equation can be reduced to:

                           s 
                  λ ρ +      
                           ur 
        w=                     γ                                                                                                   (41)
                  π V (1 − br )

In this wage equation, λ is the parameter measuring the bargaining power of workers, ρ is the
discount rate, s the (exogenous) separation rate, πV the probability of filling a vacancy and ur
the rate of unemployment. The cost of an open vacancy for the firm is measured by γ and br is
the wage replacement rate of the unemployment benefit as defined in (4). The separation rate
could be endogenized in the labour demand block and usually depends on workers
productivity, like in Faia, et al. (2013). As Boeters and Savard (2013) point out, some of these
variables are difficult to measure or derive from official data. One important property of the
wage function is the reaction of the wage rate to the unemployment rate, which according to
the empirical 'wage curve' literature is about -0.1. Taking these theoretical considerations as a
starting point we derive the following log-linearized wage curve by industry:
                                                t                              t
       log w j ,t (1 − t L ,t ) = α w, j +     ∑β        1,wj   log pct +    ∑β       2 ,wj   log(Q j ,t / H j ,t ) +
                                             t =t −n                        t =t −n
              t                                     t                                         t
       +   ∑β
           t =t −n
                     3,wj   log(Qt / H t ) +   ∑β
                                               t =t −n
                                                          4 ,wj   log(ur * / ur t ) −     ∑β
                                                                                          t =t −n
                                                                                                    5,wj   log( H j ,t / L j ,t )   (42)

The specification in (42) takes into account different lags of variables, including the consumer
price, and the industry (j) productivity or alternatively the aggregate productivity of the
economy. The term ur*/urt considers the unemployment elasticity of the wage rate in terms of
the difference to the equilibrium rate ur*, measured in that case as the minimum rate in the
sample used for estimation. The estimation of the parameter β4j,w yields the same result (only
with β4j,w > 0) as the parameter of the unemployment rate elasticity in the traditional wage
curve, because all the variance in the term ur*/urt stems from changes in the unemployment
rate. The specification of the unemployment term as a gap to full employment (ur*/urt) yields
20

a NAWRU characteristic: wage inflation increases with approximation to full employment.
Due to non-stationarity of the variables, an autoregressive term is also included. The
separation rate and the probability of filling a vacancy have not been included into (42) due to
data availability and the income replacement rate of the unemployment benefit did not yield
significant results in the panel data estimation across European countries. The stylized facts
on the latter phenomenon reveal that there is no clear correlation between the generosity of
the unemployment benefit regulation and the unemployment rate.
Labour supply is given by age and gender (g) specific participation rates of the k age groups
of the population at working age (16-65) and evolves over time according to demographic
change (age group composition) and logistic trends of the participation rates. Therefore,
labour supply does not react endogenously to policy shocks. Unemployed persons are the
difference between labour supply and employment, for given hours worked per person:

                                                L 1
       UN t = ∑ π k , g ,t popk , g ,t −wt H t  t                                      (43)
              k ,g                               H  w
                                                 t t

Total wages are given in analogy to the other factor inputs (E, Mm, and Md) by multiplying the
diagonal matrix of the factor shares in (27), in that case V̂L , with the column vector of output

                                                                                 [
                                                                                ˆ p Q
in current prices and summing up (with i' as the summation vector): wt H t = i' VD,t Q,t t .]

1.4 Government and model closure
The public sector balances close the model and show the main interactions between
households, firms and the general government. As we put special emphasis on labour market
policies, unemployment benefits are separated from the other social expenditure categories.
Taxes from households and firms are endogenized via tax rates and the path of the deficit per
GDP share according to the EU stability programs is included as a restriction.
Wage income of households is taxed with social security contributions (tax rates twL and tL)
and wage income plus operating surplus accruing to households are taxed with income taxes
(tax rate tY). Additionally, households’ gross profit income is taxed with tax rate tr. Taxes less
subsidies are not only levied on private consumption as described in (25), but also on the
other final demand components in purchaser prices (fpp, comprising capital formation,
21

changes in stocks, exports, and public consumption) as well as on gross output. Total tax
revenues of government, Tt, are given with:

           Tt = (t wL + t L )wt H t + tY ( wt H t + Π h ,t ) + tr rt At −1 + TN pp,t   [
                                                                             ˆ c +f +p Q
                                                                                     pp,t Q,t t         ]                  (44)
Taxes less subsidies and profit income in (44) also include the economic activity of the public
sector itself. The expenditure side of government is made up of unemployment transfers
(brwt(1 - tS - tY)UNt) and other transfers to households (Tr), public investment (cfgov) and
public consumption (cg). Additionally, the government pays interest with interest rate rgov on
the stock of public debt, Dgov. The change in this public debt is equal to negative government
net lending, which is then given by:

∆Dgov ,t = brwt (1 − t S − tY )UN t + Trt + cf gov ,t + cg t + rgov ,t Dgov ,t −1 − (t wL + t L )wt H t − tY ( wt H t + Π h ,t )
                    [
− tr rt At −1 − Tˆ N c pp,t + f pp,t + p Q,t Q t   ]
                                                                                                                           (45)
In that specification, tax revenues and unemployment benefits are endogenous and can from a
policy perspective be influenced by changing tax rates or the unemployment benefit
replacement rate. The model is closed by further introducing a public budget constraint,
specified via the stability program for public finances of Spain that defines the future path of
government net lending to GDP (pyY). The latter can be defined as the difference between
total output pQQ and intermediate demand (pEE, pMmMm, pMdMd). Linking public investment
with a fixed ratio (wcf) to public consumption and introducing the net lending to GDP
constraint, public consumption is then derived as the endogenous variable that closes the
model:

           cg (1 + wcf ) = ∆Dgov ,t / p yY − rgov ,t Dgov ,t −1 − brwt (1 − t S − tY )UN t − Tr + (t wL + t L )wt H t
                                                        [
           + tY ( wt H t + Π h ,t ) + tr rt At −1 + Tˆ N c pp,t + f pp,t + p Q,t Q t   ]                                   (46)

Therefore, transfers and tax rates are treated like fiscal policy variables, whereas public
consumption and investment adjust according to the net lending to GDP constraint. Public
investment can be still treated as a policy variable, as the public investment ratio (wcf) could
be altered.

2. Data, estimation and calibration
22

The data for the estimation of consumption demand functions are mainly taken from
EUROSTAT’s National Accounts. That comprises the expenditure data as well as all income
components and asset data, which are part of cash on hand. The categories of durable
consumption in our model comprise investment in own houses and purchases of vehicles. Due
to the specific treatment of housing in the consumption accounts of national accounting,
investment in own houses is pooled together with other dwelling investment to derive total
dwelling investment. In a first step, a capital stock of housing property was estimated for one
year, based on the Household Financial and Consumption Survey (HFCS) of the ECB. By
applying property prices from the Bank of International Settlement (BIS) and EUROSTAT
population data, a time series of own houses was constructed for those 14 EU countries where
sectoral accounts (income, asset data) were available from 1995 to 2011. A more simple
procedure could be applied to vehicles, as the expenditure data are available (Cveh in (47)) and
no re-valuation of the existing stock needed to be taken into account there. For own houses,
the dwelling investment (CFhous) was calculated as implicit. Measuring all variables in current
prices, the two capital stocks Kveh and Khous in current prices accumulate according to the
following equations:

        K veh,t = K veh,t −1 (1 − δ veh ) + Cveh,t                                        (47)

                       phous ,t
        K hous ,t =               [K hous,t−1 (1 − δ hous,t ) + CFhous,t / pCF ,t ]       (48)
                      phous ,t −1

In (48) revaluation of the stock is driven by the yearly change in house prices. The price phous
is the price of the housing stock and comprises increases in construction prices (pCF) as well
as changes in land prices. The variables Cveh and CFhous add up to total gross capital formation
by households, a variable that is also found in the sectoral accounts of households in National
Accounts. Given the demand and the accumulated stock of own houses, imputed rents are
calculated by applying a user cost formulation. These imputed rents enter the consumption
accounts. The expenditure for imputed rents, vehicles and total nondurables adds up to total
private consumption.The down payment for durable purchases, θCt is calculated by relating
the change in liabilities to the durable demands (Cveh and CFhous), that gives (1 - θCt). The
original θt from Luengo-Prado (2006) is measured in this model by the relationship (1 –
liabilities/durable stock) and can only be controlled by fixing certain values of θCt and solving
23

the model to derive the path of θt. In an iterative procedure dynamic convergence towards
target values of θt can then be achieved.

The functions for the two durable demand categories and for total nondurables have been
estimated with panel data econometrics for 14 EU countries (1995 - 2011), based on
EUROSTAT and other sources. Non-linear relationships of durable consumption and ‘cash on
hand’ have been identified from these estimations. Non-stationarity has not been considered
by normalizing by permanent income as is usual in the calibrated versions of the buffer stock
model, but by directly estimating an error correction mechanism (ECM) model. Table 1
shows the short- and long-run parameters for the two durable categories as well as for total
nondurables. The small impact of θCtlog(Cdur,t), i.e. of the need to finance the down payment,
is an indication that the consumer smoothes consumption and does not fully react to shocks in
liquidity. The model has been calibrated for the income quintiles in Spain, based on income
data from EU SILC and wealth data from the HFCS survey. Table 2 shows that in the base
year only small differences can be found in the marginal propensity of consumption with
respect to cash on hand.
The other more important property of the buffer stock model, however, is the reaction of
consumption (growth) to lagged income growth (excess sensitivity), an empirical
phenomenon first found by Hall (1978) and challenging the permanent income hypothesis.
This is usually tested by an OLS regression of consumption growth on lagged income growth,
including a constant. Luengo-Prado (2006) presents excess sensitivity results from US
stylized macroeconomic facts and confronts these results with results from her calibrated
model. The excess sensitivity coefficients found by Luengo-Prado (2006) are 0.16
(nondurables) and 0.26 (durables). For the DYNK model presented in this paper, excess
sensitivity has been tested at the level of income quintiles and based on the baseline run of the
model for Spain until 2050. Sensitivity analysis has been carried out for high and low θ, i.e.
the long-run liquidity constraint between the durable stock and household debt. In the “low θ
“ case this ratio still increases slightly and then converges to a value that is considerably
higher than the pre-crisis level in Spain. In the “high θ “ case debt de-laveraging takes place
and the ratio of debt to durables decreases to the pre-crisis level. Table 3 reveals that excess
sensitivity is only relevant for the first three quintiles, for the others no significant impact of
(lagged) income growth on consumption growth is found. Significant differences in the
24

impact can be found with values on average higher than those found in Luengo-Prado (2006).
The sensitivity to the down payment constraint can be seen clearly and the reactions of
consumption to income are expected to be higher in times of more binding liquidity
constraints.
The energy expenditure of households is based on consumption expenditure data from
EUROSTAT, the Energy Accounts from the WIOD database, as well as IEA Energy Prices.
In order to calculate service prices, energy efficiency data had to be added. Energy efficiency
for electricity is calculated as a weighted average of efficiency of electrical appliances from
the ODYSSEE database. The efficiency for heating is approximated by the indicator for
heating efficiency in the ODYSSEE database. Heat efficiency of the car fleet could in a
revised version also be taken from the database of the GAINS project. The durable stock of
households (vehicles, houses, appliances) embodies the efficiency of converting an energy
flow into a service level. Policy measures that increase the efficiency of the new durables
purchased or speed up the renovation of the durable stock by premature scrapping, therefore
lead to less direct energy demand of households and rebound effects from higher service
demand. The panel data set resulting from this data collection process comprises all EU 27
countries. Table 4 shows that the implicit price elasticities of transport demand are relatively
high. Note that due to the fact that the consumption variables for energy on the left hand side
are in current prices, the price elasticity is given by 1 minus the corresponding parameter
(γfuel, γheat, and γel). The expenditure elasticity of total transport demand is considerably below
unity and the density of vehicle endowment is an important factor that dampens demand for
driving. Table 5 also reveals relatively high price elasticities for given durable stocks and
controlling for climate conditions.
The results of the estimation of the demand system for non-energy nondurables are condensed
in Table 6. While the expenditure elasticities are closely distributed around unity, the price
elasticity shows more heterogeneity across categories. This elasticity mainly determines the
reactions to commodity taxation in consumption.
All data for the production system are derived from the WIOD (World Input Output
Database) dataset that contains World Input Output Tables (WIOT) in current and previous
year’s prices, Environmental Accounts (EA), and Socioeconomic Accounts (SEA). The latter
are used to derive data for capital and labour, like the base year capital stock and depreciation
25

rates as well as labour compensation by hour and by person. From the EA we use data of
energy use by 25 energy carriers in physical units (TJ) and CO2 emissions and combine the
physical energy inputs with information on energy prices from the IEA to get a full system of
energy quantities and prices. The WIOT in current and previous year’s prices have been used
to derive quantities and prices for (Mm) and (Md).The system of the unit cost function and the
factor cost shares has been estimated with panel data econometrics for 23 EU countries with
time series from 1995 to 2009. The systems have been estimated applying the Seemingly
Unrelated Regression (SUR) estimator for balanced panels under cross section fixed effects
for each of the 35 industries (345 observations). The estimation results yield parameter values
for all price terms which together with the factor cost shares give the own and cross price
elasticity according to the formulae for the Translog model. Table 7 to 10 contain the price
elasticities for capital, labour, energy, and imported intermediates respectively. The own price
elasticity of labour is on average about -0.5, with relatively high values in some
manufacturing industries. The own price elasticity of energy is very heterogenous across
industries and rather high in energy intensive industries.These elasticities have then in turn
been used to calibrate the production system for the DYNK model base year (2005) for Spain.
Table 11 shows the double impact of autonomous technical change on factor demand from
TFP that decreases unit costs and therefor factor inputs and from the factor bias that either
decreases or increases factor input in an industry. A great variety of elasticity values can be
observed from that.
Wage data including hours worked are taken from WIOD Sectoral Accounts and are
complemented by labour force data from EUROSTAT. The wage equations have been
estimated for the full EU 27 panel including lags of some of the independent variables as well
as of the wage rate per hour (ADL specification). Table 12 only shows the short-run
coefficients, the long-run elasticities are considerably larger and for the unemployment
elasticity (ur*/urt) the unweighted average across industries is about 0.09. The long-run
productivity elasticity of wages is also correspondingly higher and almost unity. Not all
industries show a significant impact of hours worked on the hourly wage rate. In industries
without this coefficient in the wage curve, a reduction of hours worked ceteris paribus leads
to a proportional income loss of workers.
You can also read